Many speech recognition systems use acoustic models to transcribe utterances. Typically, acoustic models provide information about the sounds or phonetic units that are included in an utterance.